[
https://issues.apache.org/jira/browse/HDFS-11294?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
James Clampffer reopened HDFS-11294:
------------------------------------
Reopening. Looks like this is a real issue that existing tests weren't
hitting. Given an HA cluster where both NN hostnames can't be resolved to any
ip(v4) addresses the rpc engine will try and dereference the first element of
an empty vector while trying to determine which node to use for failover.
Have a fix, will post soon. Want better test coverage to see if there are
similar issues elsewhere.
> libhdfs++: Segfault in HA failover if DNS lookup for both Namenodes fails
> -------------------------------------------------------------------------
>
> Key: HDFS-11294
> URL: https://issues.apache.org/jira/browse/HDFS-11294
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: hdfs-client
> Reporter: James Clampffer
> Assignee: James Clampffer
>
> Hit while doing more manual testing on HDFS-11028.
> The HANamenodeTracker takes an asio endpoint to figure out what endpoint on
> the other node to try next during a failover. This is done by passing the
> element at index 0 from endpoints (a std::vector<asio endpoint>).
> When DNS fails the endpoints vector for that node will be empty so the
> iterator returned by endpoints\[0\] is just a null pointer that gets
> dereferenced and causes a segfault.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
